Leverage is a measure used in regression to highlight observations which are outlying in the space of the predictors.
In regression models not all observations have the same influence on the final model and points which have more than most are said to be high-leverage points.
We define the leverage of the $i$th observation as
$$ h_{ii} = [\mathbf{H}]_{ii} $$
where $\mathbf{H}$ is the projection matrix
$$ \mathbf{H} = \mathbf{X}(\mathbf{X}^T\mathbf{X})^{-1}\mathbf{X}^T $$
The bounds on $h_{ii}$ are zero and unity. Points with $h_{ii} = 1$ effectively use a whole parameter to fit them.
Textbooks on regression will usually mention the topic and a full-length treatment of this and related topics is available in Cook, R D and Weisberg, S Residuals and influence in regression 1982, Chapman and Hall
